SmolLM3 3B on NVIDIA GeForce RTX 5080 · 16GB VRAM

✅ Runs comfortably — SmolLM3 3B @ Q8_0
SmolLM3 3B 3B at Q8_0 needs ~4.1 GB (weights 3.1 GB + KV 576 MB + overhead 512 MB @ 8K ctx) of your 14.4 GB usable VRAM — plenty of headroom. Expect ~110–180 tok/s (very fast).
context @ Q8_0: comfortable to 64K
quantneedsspeed
✓ Q8_04.1 GB~110–180 tok/s (very fast)
✓ Q5_K_M3.1 GB~150–250 tok/s (very fast)
✓ Q4_K_M2.8 GB~170–280 tok/s (very fast)
⚠ NVIDIA support is best-effort in v0.1 — verify before relying on it.
runlocal verdict card
download card share on x share on reddit drop it in your model-drop post
SmolLM3 3B: Runs comfortably
for your own README — links back here

Same model, other machines: Apple M1 · Apple M1 Pro · Apple M2 · Apple M3 Pro · Apple M2 Pro · Apple M4 Pro

Also runs on a NVIDIA GeForce RTX 5080: DeepSeek-R1 (distill) · Qwen2.5 · Qwen3 · Gemma 4 · Gemma 3

Best models for a NVIDIA GeForce RTX 5080 · 16GB VRAM →

Architecturally similar

These share SmolLM3 3B's KV-cache geometry — they size memory the same way, so here's how they fit on a NVIDIA GeForce RTX 5080 · 16GB VRAM too:

modelsharesverdict here
Qwen2.5-Coder 4kv/128hd ✅ Runs (tight)
StarCoder2 4kv/128hd ✅ Runs (tight)
Yi 1.5 4kv/128hd ✅ Runs comfortably
Qwen3 Coder 30B-A3B 4kv/128hd ⚠️ Partial GPU offload
Yi-Coder 9B 4kv/128hd ✅ Runs comfortably

Check any combo yourself: open the checker →